Add CodSpeed continuous performance testing

[codspeed-hq]

Summary

This PR integrates CodSpeed for continuous performance testing of the xml2arrow parsing benchmarks.

Changes

• Benchmark integration: Replaced criterion with codspeed-criterion-compat in benches/parse_benchmark.rs to enable CodSpeed's instrumentation layer. The existing Criterion API is fully preserved -- benchmarks still work with cargo bench as before.
• CI workflow: Added .github/workflows/codspeed.yml that builds and runs the parse_benchmark suite on every push to main, on pull requests, and on manual dispatch. Uses OIDC authentication and simulation mode for reproducible, hardware-independent measurements.
• README badge: Added CodSpeed badge alongside existing badges.
• Dependency: Added codspeed-criterion-compat v4.4.1 as a dev-dependency and updated Cargo.lock.

How it works

CodSpeed's simulation mode uses CPU instruction counting (via valgrind) rather than wall-clock timing, which eliminates variance from CI runner load and produces fully reproducible results. The codspeed-criterion-compat crate is a drop-in replacement for criterion that hooks into this instrumentation when running under CodSpeed, and falls back to standard Criterion behavior otherwise.

Next steps

• After merging, the first workflow run on main will establish the baseline performance data.
• Subsequent pull requests will automatically get performance reports comparing against the baseline.
• Visit codspeed.io/mluttikh/xml2arrow to view performance trends and flamegraphs.

[codspeed-hq]

Congrats! CodSpeed is installed 🎉

🆕 4 new benchmarks were detected.

You will start to see performance impacts in the reports once the benchmarks are run from your default branch.

Detected benchmarks

• 200K_measurements_5_sensors[203MB]: 3.6 s
• 100K_measurements_10_sensors[202MB]: 3.6 s
• 1K_measurements_2_sensors[413KB]: 7.3 ms
• 10K_measurements_5_sensors[10MB]: 178.9 ms

---